Edit: Turns out this topic is about amending the TOS or something, hop in 🤷

Well, let’s go with the flow. Alexander Hamilton, millions things I haven’t done, amending OGS’s TOS is for the battle 'gainst oppressors :microphone:

coughs Sorry, I meant "let’s talk about how to improve the terms of serv

I feel like you don’t need to quantify this, just like the rest of the TOS aren’t really quantified and mostly constitute rough guidelines for users and alibis for moderators – which I think is a good thing.

Yeah, it’s technically possible for a weak player to win against a much stronger player occasionally, which is why (I’m assuming you meant this) you can’t really “predict” the result of a game just based on a supposed difference in strength, but in the vein of what we’ve been saying, I really don’t think the rankings are going to be significantly affected by this. (and if so, the cause would be more the fact that the current rating system is almost surely too volatile)

So the current TOS say this:

No Cheating or Computer Help

You can NEVER use Go programs (Leela, Zen, etc.) or neural networks to analyze current ongoing games unless specifically permitted (e.g., a computer tournament). The only type of computer assistance allowed is games databases for opening lines and joseki databases for corner patterns in correspondence Go. You cannot receive ANY outside assistance on live or blitz Go games.

Here is my proposal:

About receiving assistance while playing (human or AI)

You cannot receive ANY outside assistance, including but not limited to Go software (KataGo, Leela, Zen, etc.), with the following exceptions:

UNRANKED GAMES: (1) You can receive external assistance if you have the opponent’s explicit and informed consent. “Informed” here also means they are aware of the nature of the external assistance you’re receiving. As an example, teaching games are of course allowed. (2) In correspondence games, you’re allowed to consult game databases for opening lines and joseki databases for corner patterns, even without the opponent’s consent.
RANKED GAMES: (1) In correspondence games, you’re allowed to consult game databases for opening lines and joseki databases for corner patterns. (2) Ranked teaching games are allowed only as long as the teacher wins and it’s clear that the teacher is the stronger player, so that the ranking result is “correct”. (3) Special permissions, such as a site-wide recognized AI-assisted tournament.

If you believe you know of a different situation that should be an exception, contact a moderator to ask for their opinion and permission.


  • The current title (“No Cheating or Computer Help”) doesn’t quite cover the actual extent of the topic touched in the paragraph even in its current form, which is bad design for a Terms of Service page, so I changed it.

  • I specified that teaching games are allowed, and the conditions for them to be allowed in ranked form.

  • Since there was too much info, I reorganized the whole thing, so that the reader can quickly parse it and find the answer to the specific questions they may have.

  • I kept the clause about the AI-assisted tournament, but should that tournament really be ranked?

  • The “opponent’s consent” clause should also kind of take care of this:

If the game is unranked and the opponent consents to it, I can’t really see any reason for not allowing the consulting of AIs during a game, especially correspondence games.

I’ve long thought that AI assistance can be a valuable tool for learning, and that it’s a bit of a pity that these terms of service didn’t allow it at all and just lump it in the bucket of “cheating”.

In the above proposed amendment, I assumed people would agree with this, but of course feel free to change it if people disagree.

I guess somewhere in the TOS it should say “contact a moderator or ask a question in the forum if you have doubts about these terms of service” or something like that? I wrote it for the paragraph, but I think it would be good to have it as a general principle.

Who lives, who dies, who reads your story? :microphone:

The TOS was changed slightly several years ago, but I don’t remember what the change was. You are right, I think, in your observation that the TOS is deliberately vague. I tried to get a few very focused clarifications made when I was a mod, but it went nowhere. The reason for this, I think, is that vagueness has some value. I am not going to expatiate on this; interested persons can go read the history of the U.S Constitution, in particular the debate about specifying rights. Bottom line: I don’t think this will get very far, especially if you set forth sweeping changes, as you do here.

Best thing, IMHO, is to debate one or two of the most important changes. If this works, you can move on to another, and another. Incrementalism. I haven’t even read this post; its too long and complicated. Moreover, it invites debate on such a broad range of topics that the discussion will likely descend into digressive chaos very quickly.


I understand where you are coming from since I’ve had similar problems.
In similar cases if I deem that there is something I can do, I go ahead and do it, instead of waiting for the community to “catch up” or “coordinate” or “organise” or whatever else they are keen on doing.

If they do not like it afterwards? That’s their problem and they can get off their chairs and do things better if they are so keen on them. Which they are usually not, so what you did sticks :wink:

To keep it Go specific, a lot of people disliked some of my translations of the Greek Go terms. So, I told them “fine you coordinate and come to a concensus about the correct terms and I’ll gladly fix them” and I would have honestly done that.

However, three years later they are still “coordinating” :rofl: and what translations I had made are still unchanged.

In your case now, it is very nice of you that you put it under discussion, but as you will soon find out, nothing really will come out of it and, in the end, you will do what you want to do. Next time you will probably say to yourself “been there done that” and skip the discussion and go ahead and implement your idea and have the same end result, without the added fuss.

This is a good point, but you shoud ask yourself if the new players even care about rating or if they understand how it works, even if the system was perfect.
I’ve been playing for years now and I still haven’t found the capacity to care about the math, the algorithm or the accuracy behind the rating system here. All I know is that it is not working well because other people are complaining.

Maybe just warn the new people that there are some rating issues with new players and that they might have some issues with their first games and leave it like that.

Learning to play Go is complex enough and a lot of people quit very early. No reason to burden them with extra complications with ranking.

Well, that was kinda my point :laughing: even just the procedure of

has enough details that the beginner newcomer, unless they’re particularly sharp, is likely to get wrong unless you hold their hands all the way through. Will they understand what the purpose of that exercise is? Will they understand that they do need to play the 20 random moves before resigning? Also it won’t work with just any AI bot, you need a bot that is weak enough for it to work, but they initially can’t play ranked games against bots weaker than 15 kyu because of the default 6 kyu rank. Even if they understand the 9 ranks thing you’d still have to explain that they do have a hidden rank at 6 kyu even though their profile says that they’re 12 kyu because of the Humble Rank system…

So even that procedure that appears incredibly simple to anyone who is familiar enough with the kyu ranks and OGS’s system, will often require a surprising amount of detailed explanations to ensure the beginner newcomer knows what they have to do. And this is not just a hypothesis: as I said, I tried this in the past, and AFAIU the beginner did try to do it, but did it the wrong way, so I gave up instead of trying to explain what they did wrong.

In the end, yes, for now I settled on telling beginners “For now you’ll get matched against stronger players, until you lose enough games for the system to understand your level, then you’ll be matched against people at your own level of strength; sorry, I know it probably sucks, but that’s just how the system works for now”.

But still, that doesn’t mean I can’t wonder if there is something more I can do, hence this thread :slight_smile:

Oh, and by the way:

 That’s kind of a big part of the issue and of the reason why I care about this: you don’t know what the beginners will do when they leave, because they’re not around to give their opinion anymore. But based on my experience interacting with them (and having been a beginner myself not too long ago, of course), I suspect a lot of them just give up Go itself after losing a bunch of games here, rather than going to other servers.
 Why would they imagine other servers are better, considering many of them have no idea how the kyu ranks work etc? More likely they stop coming here because they think they’re worthless at Go.


I haven’t been following everything here, but I was really surprised to see this:

So, if I, say, point out a self-atari or a joseki mistake and offer my opponent an undo, it’s considered “outside assistance”? I never would have thought to consider someone else in the same game as “outside”. Sure, it might eventually add up to “rank manipulation” but occasional help seems so different if it’s offered openly by your opponent vs. taken secretly from AI.

By the way, regarding humble rank, my impression is that it’s completely broken for matchmaking and maybe always was:

EDIT: it does at least appear to be working for the “opponent rank range” feature of automatch.


People keep saying this, but why?

Provisional players have one rating for rating calculation purposes, and a different one for matchmaking purposes. This, by itself, doesn’t “break” anything, since the rating system is perfectly capable of handling new players playing ranked games against players of different ratings, except for relatively weird situations.

The only two things I know so far that might be considered “broken” about the system are:

  1. The game thumbnails display the true rank instead of the matchmaking rank, which causes confusion.

  2. GaJ said in this thread that if he recalls correctly, sometimes the matchmaking doesn’t follow the humble rank. I have no idea what he means, and I have a suspicion that he might just be remembering point 1 incorrectly.

What is “broken” about it?

Because humble rank seems to be for display only and does not affect matchmaking at all. Is there any evidence that it does? For example at my linked comment you can see a new player with humble rank of ~12k getting multiple ranked games against dans, which should not happen under a humble matchmaking system. As another example, if you make a new account (try it!), see your 12k rank, and attempt to challenge, say, noob_bot_2 [19k] to a ranked game, it gives you an error:


Ooh, but that’s because the 9-rank restriction is following the true rank (6 kyu) instead of the humble rank (12 kyu).

Yeah, that’s a good point, I guess the 9-rank restriction should be part of the matchmaking system rather than the rating system, if it really needs to be a thing at all.

To cross the t’s and dot the i’s, I’d say that’s not really the humble rank system being broken per se, as much as the 9-rank restriction system being kinda dumb, especially for provisional players (although to be fair it does make sandbagging a bit more difficult), and the interaction between the various systems being broken. Seems unfair to put the “blame” on one of them arbitrarily :laughing:

Is there some aspect of humble rank matchmaking that you think actually works?

1 Like

Oh, right, I forgot to answer your other question: this thread was started by an 11-12 kyu player lamenting the fact that he’s forced to play against provisional players; the player in question almost exclusively plays automatch; that’s some evidence that the matchmaking system follows the humble rank.

For more evidence, you can start looking in that user’s page and skim through the profiles of the various [?] players they’ve played with, and you should see that very rarely the first ranked automatch game they played was against a 6 kyu, it’s usually against 11-12-13 kyus, in this one case even against an 18 kyu – and considering 6 kyu is literally the most common established rank on OGS, that sounds unlikely if the matchmaking system followed the true 6 kyu rank.

Admittedly I haven’t performed an extensive search, but for now this evidence is convincing enough for me.

Okay, that’s something. I almost never use automatch myself and hadn’t considered it.

I tried it out just now restricting the rank to ±1 and my new account got automatched against a 12k.

So I guess it’s not completely broken after all! Now I wonder what it will do with handicap games…


By the way, when I say “matchmaking”, I’m usually thinking exclusively about the automatch system :laughing:

For example, the example you brought in the other thread is a player playing custom computer games against a dan bot, and it didn’t really feel relevant to me, since custom games are not matchmade by the site; though it was a very useful example for other purposes.

1 Like

I wonder if this is what the word “outside” is meant to deal with?

Maybe your opponent is not “outside”?

That would solve the original question in one hit :slight_smile:

1 Like

Ok, another more simple idea. Just tell them to join the site’s ladder so now they can challenge people of the correct rank on their own. I’ve never used the matchmaker or the “play” button in this site, to be honest, and I’ve never used anything similar on DGS (Dragon Go Server).
Even as a beginner I just joined the ladder there and started choosing my opponents depending on what I wanted to do at the time (teaching game? I challenged people with higher ranking. Testing game? I challenged opponents with similar ranking.)

It is pretty simple and fun that way.


That’s how I always find matches, when I’m looking for them, too.

It allows you to select not only the rank of your opponent, but their playing characteristics, if you feel motivated to look those up (for example, I will chose an opponent who prefers resignation to playing out the dame, many folk feel precisely the opposite to this :slight_smile: ).


Well that’s a bit too much to expect from new players and I haven’t found the motivation to do that, but since you mentioned it, it reminded me of something really nice that DGS has and OGS does not have (or, at least, I haven’t noticed it), which would be very useful in this situation and topic.

DGS informs you BEFORE you issue a challenge, how much your rank and the opponent’s rank will change depending on the result (of course that is liable to change a bit by the time the correspondence game ends, but fluctuations there are minimal and who remember the exact number after 2 months, eh? :stuck_out_tongue: )

For example, if I challenge someone on my rank, I get the ranking change for both of us and since we seem to be “fresh” on our maximum ranking of 3k, the algorithm is giving us (or taking away) a sizable amount of “points” for both players.

However, if I were to challenge an established higher ranking player I would get almost half a kyu worth of points, but the dan player would only lose around 10 points, because the system knows their rank to be correct and there is no need to punish a good player for having a bad game.

In any case, a much more transparent system and with minimal looking around a player can know if their rank and their opponent’s ranking is trusted by the system and is really reflecting their strength.

1 Like

Ah right - for clarity, I didn’t mean to imply beginners would do this, I was just elaborating on some of the benefits that can flow from finding opponents this way :slight_smile:

1 Like

Ok, I think this is another blindspot that (some) established OGS players have: despite OGS culture being correspondence-oriented, correspondence is not a “natural” way to play Go, and from what I’ve seen so far, it’s also not what beginner newcomers expect when joining a Go server.

Many of the new players that do stick with correspondence are probably just adapting to the local culture, rather than following their own inclinations.

If I point the newcomers towards correspondence games, some of them may indeed stick with it, but then there's the other thing: I am genuinely not sure if playing correspondence is a good thing for most beginners.

As far as I can tell, there are two approaches to correspondence that naturally arise:

  • some players play as many simultaneous corr games as the can handle, and only spend a little time thinking of their moves;
  • some players, instead, take a loooong time to think their moves through. To me, tbh, this just sounds like the obvious way to play corr. You have as much time as you want, so why wouldn’t you make sure that you’re playing the best move?

First of all, I believe both approaches have one drawback in common: there is usually no continuity between the moves, which easily causes players, especially beginners, to be unable to keep in mind the status of the various zones of the board and follow a “flow” of the game. This especially makes reviewing the game challenging, since you probably won’t remember what you were thinking when making the moves.

 The players who don’t spend too much time thinking about their moves obviously don’t have the opportunity to read deeply into the game, which doesn’t promote learning.
 While it’s true that young beginners are able to learn from fast live games, I believe it’s also because they see their mistakes punished shortly after having made them – the correspondence usually takes away that feedback aspect.

 For the players who spend a long time thinking, there’s another problem: the temptation to consult an AI engine or generally to look for outside assistance is very high, especially for beginners, who are often confused and mystified by the game. After all, consulting an AI doesn’t feel that much different that just analyzing future variations on your own.
 So there’s an escalating behaviour that can establish where you start just consulting the AI to explore variations and then make your own move, to gradually end up just straight up cheating if you get attached to winning.

 Since many of these beginners are young, they don’t really care about the rules (and they’re probably unaware of AI consulting being against the TOS) until they’re explicitly warned or punished for them, and by that time it might be too late – their reaction to being warned that AI consulting is against the rules might just be to keep doing it, but discreetly as to not get caught.

There are several reasons why I usually don’t play corr, and the temptation to consult AIs is one of them.

So, uh, I don’t know about that :laughing:


Hey there’s a typo!


Is this an inside joke or something? :laughing:

The only “unorthodox” thing I can see there is “Can not” instead of “Cannot”, but I wouldn’t really consider it a typo :thinking: